In one paragraph
California's SB 503 targets bias in clinical AI. Developers and deployers (health facilities, clinics and medical practices) of clinical decision support systems, AI that aids decisions on timing of care, diagnosis or treatment, must make reasonable efforts to find systems with known or foreseeable risk of biased impacts based on protected characteristics. Developers must mitigate those risks and give deployers a statement of uses and risks plus documentation on training data, bias testing and monitoring; deployers must monitor the systems and take reasonable steps to mitigate bias. Signed September 30, 2026 (Chapter 857); effective January 1, 2027.
Plain-English summary, written with AI assistance. The official text controls.
What it means for business
If you build or use clinical decision-support AI in California health care, from Jan. 1, 2027 you must make reasonable efforts to identify and mitigate bias risks; developers must document training data and bias testing for their customers.
Key provisions
- New Bus. & Prof. Code § 22758 (Chapter 25.3, Clinical Decision Support Systems)
- Clinical decision support system: AI that produces a prediction, classification, recommendation, evaluation or analysis aiding clinical decisions on timing of care, diagnosis or treatment; appointment management, reminders, patient education and payment processing are excluded
- Developers and deployers make reasonable efforts to identify systems known, or reasonably foreseeable, to risk biased impacts: adverse effects on access to care, quality or outcomes based on protected characteristics (Civil Code § 51(b))
- Developers make reasonable efforts to mitigate those risks and give deployers a statement of intended uses and risks, plus documentation on training data and its demographic representativeness, performance and bias evaluation, data governance, known risks and mitigation, and recommended use and monitoring
- Documentation is due on request or at initial sale, whichever is earlier, and with material updates; developers may rely on recognized industry standards or algorithmic impact assessment results
- Deployers (health facilities, clinics, physician offices and group practices) regularly monitor identified systems and take reasonable, proportionate steps to mitigate biased impacts
- Adds to other AI and automated-decision laws; compliance is not a defense to a discrimination claim
- Signed September 30, 2026 (Chapter 857, Statutes of 2026); effective January 1, 2027
Latest legislative action
Sep 30, 2026. Approved by the Governor; Chapter 857, Statutes of 2026
Sponsors (showing 5 of 52)
| Name | Role |
|---|---|
| Alpert | Principal Coauthor |
| Alvarado-Gil | Author |
| Archie-Hudson | Coauthor |
| Ayala | Author |
| Becker | Author |
Roll call votes (showing 3 of 8)
Do pass. Aug 29, 2025
15 yea0 nayPassed
Do pass as amended and be re-referred to the Committee on [Appropriations] Jul 16, 2025
13 yea0 nay2 otherPassed
Do pass as amended and be re-referred to the Committee on [Privacy and Consumer Protection] Jul 8, 2025
16 yea0 nayPassed
Topics and industries
Related topics
Affected industries
This page is general information, not legal advice. Summaries are AI-assisted and may contain errors; read the statute for the controlling text and talk to a lawyer about your situation. Full disclaimer · Report an error